/*
* Document-method: parse
*
* call-seq:
* parse(input, buffer_size=8092)
* parse(input, buffer_size=8092) { |obj| ... }
*
* +input+ can either be a string or an IO to parse JSON from
*
* +buffer_size+ is the size of chunk that will be parsed off the input (if it's an IO) for each loop of the parsing process.
* 8092 is a good balance between the different types of streams (off disk, off a socket, etc...), but this option
* is here so the caller can better tune their parsing depending on the type of stream being passed.
* A larger read buffer will perform better for files off disk, where as a smaller size may be more efficient for
* reading off of a socket directly.
*
* If a block was passed, it's called when an object has been parsed off the stream. This is especially
* usefull when parsing a stream of multiple JSON objects.
*
* NOTE: you can optionally assign the +on_parse_complete+ callback, and it will be called the same way the optional
* block is for this method.
*/
static VALUE rb_yajl_parser_parse(int argc, VALUE * argv, VALUE self) {
yajl_status stat;
yajl_parser_wrapper * wrapper;
VALUE rbufsize, input, blk;
unsigned int len;
const char * cptr;
GetParser(self, wrapper);
/* setup our parameters */
rb_scan_args(argc, argv, "11&", &input, &rbufsize, &blk);
if (NIL_P(rbufsize)) {
rbufsize = INT2FIX(READ_BUFSIZE);
} else {
Check_Type(rbufsize, T_FIXNUM);
}
if (!NIL_P(blk)) {
rb_yajl_parser_set_complete_cb(self, blk);
}
if (TYPE(input) == T_STRING) {
cptr = RSTRING_PTR(input);
len = RSTRING_LEN(input);
yajl_parse_chunk((const unsigned char*)cptr, len, wrapper->parser);
} else if (rb_respond_to(input, intern_io_read)) {
VALUE parsed = rb_str_new(0, FIX2LONG(rbufsize));
while (rb_funcall(input, intern_io_read, 2, rbufsize, parsed) != Qnil) {
cptr = RSTRING_PTR(parsed);
len = RSTRING_LEN(parsed);
yajl_parse_chunk((const unsigned char*)cptr, len, wrapper->parser);
}
} else {
rb_raise(cParseError, "input must be a string or IO");
}
/* parse any remaining buffered data */
stat = yajl_parse_complete(wrapper->parser);
if (wrapper->parse_complete_callback != Qnil) {
yajl_check_and_fire_callback((void *)self);
return Qnil;
}
// Because we've overridden the yajl_parse_complete state of the parser to allow
// us to parse multiple objects out of one stream; we can't use it to determine
// whether we've parsed a single object. Instead we'll check whether we've successfully
// parsed a single token.
if (!RARRAY_LEN(wrapper->builderStack) ||
wrapper->nestedHashLevel ||
wrapper->nestedArrayLevel) {
rb_raise(cParseError, "unexpected end of JSON string");
}
return rb_ary_pop(wrapper->builderStack);
}
/*
* Document-method: parse
*
* call-seq:
* parse(input, buffer_size=8092)
* parse(input, buffer_size=8092) { |obj| ... }
*
* +input+ can either be a string or an IO to parse JSON from
*
* +buffer_size+ is the size of chunk that will be parsed off the input (if it's an IO) for each loop of the parsing process.
* 8092 is a good balance between the different types of streams (off disk, off a socket, etc...), but this option
* is here so the caller can better tune their parsing depending on the type of stream being passed.
* A larger read buffer will perform better for files off disk, where as a smaller size may be more efficient for
* reading off of a socket directly.
*
* If a block was passed, it's called when an object has been parsed off the stream. This is especially
* usefull when parsing a stream of multiple JSON objects.
*
* NOTE: you can optionally assign the +on_parse_complete+ callback, and it will be called the same way the optional
* block is for this method.
*/
static VALUE rb_yajl_parser_parse(int argc, VALUE * argv, VALUE self) {
yajl_status stat;
yajl_parser_wrapper * wrapper;
VALUE rbufsize, input, blk;
unsigned int len;
const char * cptr;
GetParser(self, wrapper);
/* setup our parameters */
rb_scan_args(argc, argv, "11&", &input, &rbufsize, &blk);
if (NIL_P(rbufsize)) {
rbufsize = INT2FIX(READ_BUFSIZE);
} else {
Check_Type(rbufsize, T_FIXNUM);
}
if (!NIL_P(blk)) {
rb_yajl_parser_set_complete_cb(self, blk);
}
if (TYPE(input) == T_STRING) {
cptr = RSTRING_PTR(input);
len = RSTRING_LEN(input);
yajl_parse_chunk((const unsigned char*)cptr, len, wrapper->parser);
} else if (rb_respond_to(input, intern_io_read)) {
VALUE parsed = rb_str_new(0, FIX2LONG(rbufsize));
while (rb_funcall(input, intern_io_read, 2, rbufsize, parsed) != Qnil) {
cptr = RSTRING_PTR(parsed);
len = RSTRING_LEN(parsed);
yajl_parse_chunk((const unsigned char*)cptr, len, wrapper->parser);
}
} else {
rb_raise(cParseError, "input must be a string or IO");
}
/* parse any remaining buffered data */
stat = yajl_parse_complete(wrapper->parser);
if (wrapper->parse_complete_callback != Qnil) {
yajl_check_and_fire_callback((void *)self);
return Qnil;
}
return rb_ary_pop(wrapper->builderStack);
}
/*
* Document-method: parse_chunk
*
* call-seq: parse_chunk(string_chunk)
*
* +string_chunk+ can be a partial or full JSON string to push on the parser.
*
* This method will throw an exception if the +on_parse_complete+ callback hasn't been assigned yet.
* The +on_parse_complete+ callback assignment is required so the user can handle objects that have been
* parsed off the stream as they're found.
*/
static VALUE rb_yajl_parser_parse_chunk(VALUE self, VALUE chunk) {
yajl_parser_wrapper * wrapper;
unsigned int len;
GetParser(self, wrapper);
if (NIL_P(chunk)) {
rb_raise(cParseError, "Can't parse a nil string.");
}
if (wrapper->parse_complete_callback != Qnil) {
const char * cptr = RSTRING_PTR(chunk);
len = RSTRING_LEN(chunk);
yajl_parse_chunk((const unsigned char*)cptr, len, wrapper->parser);
} else {
rb_raise(cParseError, "The on_parse_complete callback isn't setup, parsing useless.");
}
return Qnil;
}
